Methodology Matters: A Careful Meta-Analysis of Climate Damages
通过更新数据和方法，重新估计了全球变暖3°C时的非灾难性损害占GDP的3.2%-9.2%，并发现与先前研究的分歧主要源于数据选择和方法控制差异。
Abstract Environmental economists widely use meta-regression to combine estimates from multiple studies to strive for consensus Nelson and Kennedy (Environ Resour Econ 42(3):345–77, 2009). Despite its wide-scale adoption, climate economists using meta-regression to synthesize climate damage estimates find a wide range of outcomes when analyzing impacts at the global scale, as represented by the primary and alternative damage functions in Barrage and Nordhaus (Proc Natl Acad Sci USA 121(13):e2312030121, 2024). Conducting a new search and selection procedure, we update the underlying data from Howard and Sterner (Environ Resour Econ 68(1)197–225, 2017), the study which represents the upper end of this range. Using a slightly modified version of the Howard and Sterner (Environ Resour Econ 68(1)197–225, 2017) model enabled by our larger dataset, we estimate non-catastrophic damages to be between 3.2% and 9.2% of GDP for a 3 °C increase depending on whether growth effects are included. This range increases to 12.5% and 18.5% with the inclusion of catastrophic impacts. Using probabilistic weights for these uncertain structural assumptions, our best estimate is between 7.1% and 12.6% depending on how narrowly we define catastrophic damages. We explore the causes of the disparity between our results and Barrage and Nordhaus (Proc Natl Acad Sci USA 121(13):e2312030121, 2024), and find that it stems primarily from differences in data search, selection, and weighting (with our study including more independent studies and a wider set of estimation strategies) and our decision to include methodological controls. To resolve the disparity, it is critical to adjust for methodological differences between studies, as the underlying mix of estimation methodologies is non-random and may capture different, but overlapping, impacts of climate change.